Road traffic natural disaster early warning, prevention and control system based on AIOT

Abstract

The invention relates to the field of natural disaster early warning, in particular to a road traffic natural disaster early warning, prevention and control system based on AIOT. The technical problems are that a road traffic natural disaster early warning prevention and control system in the prior art often depends on manual judgment or simple threshold setting, is not intelligent enough, mostly depends on a single data source for analysis, and ignores other important factors such as meteorology and geography; according to the technical scheme, the road traffic natural disaster early warning, prevention and control system based on the AIOT comprises an edge layer module, a transmission layer module, a data layer module, a supporting layer module and an application layer module; multi-dimensional data such as temperature and humidity, wind power and wind direction, rainfall, PM 2.5-10 and video images are fused, deep learning and mining are carried out through the AI intelligent analysis unit, the accuracy and comprehensiveness of natural disaster early warning are improved, automatic intelligent early warning and alarming are achieved through logical judgment and an AI algorithm, man-made misjudgment is reduced, and the working efficiency is improved. And the timeliness and accuracy of early warning are improved.

Description

Technical Field

[0001] The present invention relates to the field of natural disaster early warning, and in particular to an AIOT-based road traffic natural disaster early warning and prevention system. Background Art

[0002] The road traffic natural disaster early warning and prevention system is a comprehensive system that integrates monitoring, analysis, early warning and prevention and control. It aims to improve the response capability and safety of road traffic in the face of natural disasters. The system can automatically transmit early warning information to relevant construction units and construction units, activate emergency plans, and realize the closed-loop management of the entire process of early warning response and disposal; this system effectively reduces the impact of natural disasters on road traffic and protects the lives and property of the people; while the road traffic natural disaster early warning and prevention system under the existing technology often relies on manual judgment or simple threshold setting, is not intelligent enough, and mostly relies on a single data source for analysis, ignoring other important factors such as meteorology and geography. Summary of the invention

[0003] In order to overcome the problem that road traffic natural disaster early warning and prevention systems under existing technologies often rely on manual judgment or simple threshold settings, are not intelligent enough, and mostly rely on a single data source for analysis, ignoring other important factors such as meteorology and geography.

[0004] The technical solution of the present invention is: a road traffic natural disaster early warning and prevention system based on AIOT, comprising:

[0005] The edge layer module is used to collect road weather and natural disaster data in real time, perform local preprocessing and logical judgment through edge computing strategies, and provide preliminary warning and alarm for abnormal conditions;

[0006] The transport layer module is used to transmit the data collected by the edge layer devices to the cloud server;

[0007] The data layer module is used to store, clean, analyze and mine the received data and provide data support for upper-layer applications;

[0008] The support layer module is used to provide stable, secure and efficient technical support for the application layer to ensure the efficient operation of the system;

[0009] The application layer module is used to realize real-time monitoring, early warning, alarm and data analysis of natural disasters based on the data provided by the data layer, and provide decision-making support for traffic safety supervision;

[0010] The security system layer module is used to adopt traffic safety laws, regulations and technical specifications system and information security management and information system level protection system to ensure that system construction complies with national and local laws and regulations and meets national information security requirements.

[0011] Preferably, the edge layer module is used to collect road weather and natural disaster data in real time, and local preprocessing and logical judgment are performed through edge computing strategies to carry out preliminary early warning and alarm of abnormal conditions; the data collected by the edge layer equipment is transmitted to the cloud server through the transport layer module; the received data is stored, cleaned, analyzed and mined through the data layer module to provide data support for upper-level applications; the support layer module provides stable, secure and efficient technical support for the application layer to ensure efficient operation of the system; the application layer module realizes real-time monitoring, early warning, alarm and data analysis of natural disasters based on the data provided by the data layer, and provides decision-making support for traffic safety supervision; the security system layer module adopts traffic safety laws, regulations and technical specification system and information security management and information system level protection system to ensure that the system construction complies with national and local laws and regulations and meets national information security requirements.

[0012] Preferably, the edge layer module includes temperature and humidity sensors, wind speed and direction sensors, rainfall sensors, PM2.5-10 sensors, video cameras and natural disaster intelligent monitoring terminals; each sensor and video camera collects road weather and natural disaster data in real time, and the natural disaster intelligent monitoring terminal performs preliminary processing on the collected data through edge computing strategies, and determines whether there is an abnormal state of natural disasters based on preset thresholds and logical judgment rules, and realizes early warning and alarm of abnormal states of the local natural environment through edge computing control strategies.

[0013] Preferably, when the edge layer module is working, the following steps are included:

[0014] S101: Various sensors and video cameras collect road weather and natural disaster data in real time;

[0015] S102: The natural disaster intelligent monitoring terminal performs preliminary data cleaning to remove noise data, abnormal values ​​or duplicate data;

[0016] S103: Using edge computing strategy, the saturated water vapor pressure is calculated using the formula recommended by Emanuel and the Tetens formula. The formula recommended by Emanuel provided by the International Meteorological Organization is as follows:

[0017] Among them, E1 is the saturated water vapor pressure, the unit is hPa (hectopascal); T is the absolute temperature scale, and the relationship with Celsius temperature t (℃) is: T = t + 273.15, that is, the relationship between saturated water vapor pressure and Celsius temperature t is as follows:

[0018]

[0019] There are two formulas for calculating the saturated water vapor pressure of Tetens, one for the water surface and the other for the ice surface, which are:

[0020] Water surface E * = 6.11×10 7.5t / 237.3+t 、Ice surface E * = 6.11×10 9.5t / 265.5+t ;

[0021] S104: After modifying the Tetens saturated water vapor pressure calculation formula, the formula is as follows:

[0022] Where E2 is the saturated water vapor pressure; t is in degrees Celsius;

[0023] S105: Arithmetically average the calculation formulas in steps S104 and S103 to obtain the arithmetic mean of the saturated water vapor pressure. The calculation formula is as follows:

[0024]

[0025] S106: Calculate the relative humidity according to the definition that the relative humidity is the ratio of the actual water vapor pressure in the air to the saturated water vapor pressure at the same temperature. The calculation formula is as follows:

[0026] Where E is the actual water vapor pressure in the air at the same temperature, and e is the saturated water vapor pressure at the same temperature;

[0027] S107: Calculate the actual water vapor pressure according to the relative humidity calculation formula. The calculation formula is as follows:

[0028] e = f×E; where E is the saturated water vapor pressure at the current humidity;

[0029] S108: Calculate the dew point temperature. The calculation formula is as follows:

[0030] In the same meteorological environment, when E = e,

[0031] T1 = t; where T1 is the dew point temperature and t is the monitored temperature of the temperature sensor;

[0032] S109: According to the above algorithm derivation, when e = E and t < T1, fog banks form on the road; if the MP2.5 - 10 value is greater than the threshold, haze forms on the road; if the air temperature continues to drop below 0°C, frost and ice form on the road;

[0033] S110: By comparing the saturated water vapor pressure with the humidity, determine whether the road humidity reaches saturation. If it reaches saturation, the temperature at this time is the dew temperature. If the temperature continues to drop, the water vapor in the air will condense to form "fog". If the PM value is greater than the threshold, "haze" forms; if the temperature still drops to 0°C, frost and ice will form on the road surface;

[0034] S111: When it is determined that an abnormal state exists, a real-time warning is issued through a local sound and light alarm device;

[0035] S112: Upload the collected data and logical judgment results to the cloud.

[0036] Preferably, the transport layer module receives data packets from the edge layer module, selects the optimal transmission path according to the destination address of the data packets and the current network environment, and transmits the data packets to the cloud using the project's dedicated network and 4G / 5G dedicated wireless network.

[0037] Preferably, the data layer module includes a natural disaster perception data unit, a video surveillance data unit, a traffic indication status data unit, a traffic safety business supervision data unit and an AI intelligent analysis unit; the data layer module receives data packets from the transport layer module, parses the data packets, and extracts meteorological and natural disaster data; stores the data in a MySQL database, and performs data cleaning and preprocessing. The AI ​​intelligent analysis unit uses a multi-dimensional natural disaster convolutional neural network algorithm to perform deep learning and mining on big data to analyze the trends and patterns of natural disasters.

[0038] Preferably, the data layer module includes the following steps when performing natural disaster data analysis:

[0039] S201: resize the video frame to a uniform size of 96x96 pixels to match the input requirements of the CNN model;

[0040] S202: Convert the image from RGB color space to grayscale;

[0041] S203: Perform normalization processing to scale the pixel value to between 0 and 1;

[0042] S204: labeling images, assigning one or more labels to each image, indicating whether there is a natural disaster and its type;

[0043] S205: aligning the meteorological data, the road geographical environment data and the traffic flow data according to the timestamps;

[0044] S206: Encode the non-numeric data according to the road type and vehicle type, and convert them into a numerical form that can be processed by the model;

[0045] S207: Convert the data into time series and feature vectors;

[0046] S208: Input image data, the input layer receives the pre-processed image frame;

[0047] S209: inputting non-image data, receiving the non-image data through another input layer;

[0048] S210: performing a convolution operation on the input image using multiple convolution kernels to extract features at different levels;

[0049] S211: Each convolution kernel is trained based on natural disaster characteristics;

[0050] S212: Apply ReLU activation function to increase nonlinearity;

[0051] S213: Add a pooling layer after the convolutional layer to reduce the size of the feature map and the amount of computation;

[0052] S214: combining the image data and the non-image data by concatenation and weighted summation using a fusion layer;

[0053] S215: Flatten the feature maps output by the convolutional layer, pooling layer, and fusion layer, and input them into the fully connected layer;

[0054] S216: The fully connected layer is used to map the extracted features to the target category and output the predicted probability of each category;

[0055] S217: The output layer uses the softmax activation function to output the predicted probability of each natural disaster category;

[0056] As a preferred embodiment, the data layer module also includes the following steps when performing natural disaster data analysis:

[0057] S218: Use the cross entropy loss function to measure the difference between the model prediction and the actual label;

[0058] S219: Select the Adam optimization algorithm to update the model weights and minimize the loss function;

[0059] S220: Divide the labeled data set into a training set, a validation set, and a test set;

[0060] S221: Train the model on the training set and update the weights through back propagation algorithm and gradient descent method;

[0061] S222: Evaluate model performance on the validation set, adjust learning rate, batch size, number of convolution kernels, and optimize the model;

[0062] S223: Use the test set to evaluate the model's accuracy, recall, and F1 score indicators;

[0063] S224: The maximum predicted value of each result obtained in S216 is taken as the recognition result, and multiple fully connected layers output the recognition results to form the output result of road natural disaster classification recognition.

[0064] Preferably, the support layer module includes a GIS electronic map, a microservice component, an MQ message queue, a load balancing service unit, a data cache service unit and a security management service unit; the GIS electronic map is used to provide geographic information services for the application layer, and supports map display and path planning; the microservice component is used to split the system into multiple microservices, each of which runs independently; the MQ message queue is used to realize asynchronous communication between modules within the system and improve the concurrent processing capability of the system; the load balancing service unit is used to dynamically allocate requests to different servers according to the system load to achieve load balancing; the data cache service unit is used to cache commonly used data, reduce the access pressure of the database, and improve the response speed of the system; the security management service unit is used to provide security services such as identity authentication, authority management and data encryption.

[0065] Preferably, the application layer modules include a natural disaster real-time monitoring application, a natural disaster early warning alarm application, a natural environment data analysis application and a traffic safety supervision business application; the natural disaster real-time monitoring application is used to display road weather and natural disaster data in real time, and supports map display and data query; the natural disaster early warning alarm application is used to automatically generate early warning alarm information based on the data provided by the data layer and the preset threshold, and publish it through sound and light alarm devices, LED fluorescent display boards, and broadcasting; the natural environment data analysis application is used to conduct in-depth analysis of big data, explore the laws and trends of natural disasters, and provide data support for traffic safety supervision; the traffic safety supervision business application is used to provide decision support for road traffic control or driving induction for the public security traffic police based on the data analysis results.

[0066] Preferably, the security system layer module includes a traffic safety laws, regulations and technical specification system and an information security management and information system level protection system; the traffic safety laws, regulations and technical specification system is used to ensure that the system construction complies with the requirements of national and local laws, regulations and technical specifications; the information security management and information system level protection system ensures the information security and stable operation of the system by formulating information security management systems, strengthening security audits and risk assessments and other measures.

[0067] Beneficial effects of the present invention:

[0068] 1. Compared with the existing road traffic natural disaster early warning and prevention systems, which often rely on manual judgment or simple threshold setting, they are not intelligent enough and mostly rely on a single data source for analysis, ignoring other important factors such as meteorology and geography. This system integrates multi-dimensional data such as temperature and humidity, wind speed and direction, rainfall, PM2.5-10, video images, etc., and conducts deep learning and mining through AI intelligent analysis units, thereby improving the accuracy and comprehensiveness of natural disaster early warnings. Through logical judgment and AI algorithms, it realizes automatic intelligent early warning and alarm, reduces human misjudgment, and improves the timeliness and accuracy of early warnings.

[0069] 2. Use the data layer module to adjust the video frame to a uniform size of 96x96 pixels to match the input requirements of the CNN model; align meteorological data, road geographic environment data, and traffic flow data by timestamp, encode non-numeric data by road type and vehicle type, and convert them into numerical forms that the model can process; convert the data into time series and feature vectors; use multiple convolution kernels to perform convolution operations on the input image to extract features at different levels; each convolution kernel is trained for natural disaster features; apply the ReLU activation function to increase nonlinearity; add a pooling layer after the convolution layer to reduce the size and computation of the feature map; use the fusion layer to combine image data and non-image data by splicing and weighted summation The data is combined; the feature maps output by the convolution layer, pooling layer and fusion layer are flattened and input into the fully connected layer; the fully connected layer is used to map the extracted features to the target category and output the predicted probability of each category; the output layer uses the softmax activation function to output the predicted probability of each natural disaster category; the largest predicted value of each result is taken as the recognition result, and multiple fully connected layers output the recognition results to form the output result of road natural disaster classification recognition; thus, by integrating multi-dimensional data such as temperature and humidity, wind speed and direction, rainfall, PM2.5-10, video images, etc., and performing deep learning and mining through the AI ​​intelligent analysis unit, the accuracy and comprehensiveness of natural disaster warning are improved;

[0070] 3. By comparing the saturated water vapor pressure with the humidity, it is determined whether the road humidity has reached saturation. If it has reached saturation, the temperature at this time is the dew temperature. If the temperature continues to drop, the water vapor in the air will be condensed to form "fog". If the PM value is greater than the threshold, "haze" will be formed. At this time, if the temperature continues to drop to 0°C, frost and ice will form on the road surface. When it is determined that there is an abnormal state, a real-time warning is issued through the local sound and light alarm device. Thus, through logical judgment and AI algorithms, automatic intelligent warning and alarm are realized, which reduces human misjudgment and improves the timeliness and accuracy of the warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1Shown is a schematic diagram of the overall architecture of the road traffic natural disaster early warning and prevention system based on AIOT of the present invention;

[0072] Figure 2 What is shown is a schematic diagram of the working process principle of the road traffic natural disaster early warning and prevention system based on AIOT of the present invention. DETAILED DESCRIPTION

[0073] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0074] See also Figure 1-2 The present invention provides an embodiment: a road traffic natural disaster early warning and prevention system based on AIOT, comprising:

[0075] The edge layer module is used to collect road weather and natural disaster data in real time, perform local preprocessing and logical judgment through edge computing strategies, and provide preliminary warnings and alarms for abnormal conditions;

[0076] The transport layer module is used to transmit the data collected by the edge layer devices to the cloud server;

[0077] The data layer module is used to store, clean, analyze and mine the received data and provide data support for upper-layer applications;

[0078] The support layer module is used to provide stable, secure and efficient technical support for the application layer to ensure the efficient operation of the system;

[0079] The application layer module is used to realize real-time monitoring, early warning, alarm and data analysis of natural disasters based on the data provided by the data layer, and provide decision support for traffic safety supervision;

[0080] The security system layer module is used to adopt traffic safety laws, regulations and technical specifications system and information security management and information system level protection system to ensure that system construction complies with national and local laws and regulations and meets national information security requirements.

[0081] Preferably, the edge layer module is used to collect road weather and natural disaster data in real time, and local preprocessing and logical judgment are performed through edge computing strategies to carry out preliminary early warning and alarm of abnormal conditions; the data collected by the edge layer equipment is transmitted to the cloud server through the transport layer module; the received data is stored, cleaned, analyzed and mined through the data layer module to provide data support for upper-level applications; the support layer module provides stable, secure and efficient technical support for the application layer to ensure efficient operation of the system; the application layer module realizes real-time monitoring, early warning, alarm and data analysis of natural disasters based on the data provided by the data layer, and provides decision-making support for traffic safety supervision; the security system layer module adopts traffic safety laws, regulations and technical specification system and information security management and information system level protection system to ensure that the system construction complies with national and local laws and regulations and meets national information security requirements.

[0082] Preferably, the edge layer module includes temperature and humidity sensors, wind speed and direction sensors, rainfall sensors, PM2.5-10 sensors, video cameras and natural disaster intelligent monitoring terminals; each sensor and video camera collects road weather and natural disaster data in real time, and the natural disaster intelligent monitoring terminal performs preliminary processing on the collected data through edge computing strategies, and determines whether there is an abnormal state of natural disasters based on preset thresholds and logical judgment rules, and realizes early warning and alarm of abnormal states of the local natural environment through edge computing control strategies.

[0083] Preferably, when the edge layer module is working, the following steps are included:

[0084] S101: Various sensors and video cameras collect road weather and natural disaster data in real time;

[0085] S102: The natural disaster intelligent monitoring terminal performs preliminary data cleaning to remove noise data, abnormal values ​​or duplicate data;

[0086] S103: Using edge computing strategy, the saturated water vapor pressure is calculated using the formula recommended by Emanuel and the Tetens formula. The formula recommended by Emanuel provided by the International Meteorological Organization is as follows:

[0087] Among them, E1 is the saturated water vapor pressure, the unit is hPa (hectopascal); T is the absolute temperature scale, and the relationship with Celsius temperature t (℃) is: T = t + 273.15, that is, the relationship between saturated water vapor pressure and Celsius temperature t is as follows:

[0088]

[0089] There are two formulas for calculating the saturated water vapor pressure of Tetens, one for the water surface and the other for the ice surface, which are:

[0090] Water surface E * = 6.11×10 7.5t / 237.3+t 、Ice surface E * = 6.11×10 9.5t / 265.5+t ;

[0091] S104: After modifying the Tetens saturated water vapor pressure calculation formula, the formula is as follows:

[0092] Where E2 is the saturated water vapor pressure; t is in degrees Celsius;

[0093] S105: Arithmetically average the calculation formulas in steps S104 and S103 to obtain the arithmetic mean of the saturated water vapor pressure. The calculation formula is as follows:

[0094]

[0095] S106: Calculate the relative humidity according to the definition that the relative humidity is the ratio of the actual water vapor pressure in the air to the saturated water vapor pressure at the same temperature. The calculation formula is as follows:

[0096] Where E is the actual water vapor pressure in the air at the same temperature, and e is the saturated water vapor pressure at the same temperature;

[0097] S107: Calculate the actual water vapor pressure according to the relative humidity calculation formula. The calculation formula is as follows:

[0098] e = f×E; where E is the saturated water vapor pressure at the current humidity;

[0099] S108: Calculate the dew point temperature. The calculation formula is as follows:

[0100] In the same meteorological environment, when E = e,

[0101] T1 = t; where T1 is the dew point temperature and t is the monitored temperature of the temperature sensor;

[0102] S109: According to the above algorithm derivation, when e = E and t < T1, fog banks form on the road; if the MP2.5 - 10 value is greater than the threshold, haze forms on the road; if the air temperature continues to drop below 0°C, frost and ice form on the road;

[0103] S110: By comparing the saturated water vapor pressure with the humidity, determine whether the road humidity reaches saturation. If it reaches saturation, the temperature at this time is the dew temperature. If the temperature continues to drop, the water vapor in the air will condense to form "fog". If the PM value is greater than the threshold, "haze" forms; at this time, if the temperature still drops below 0°C, frost and ice will form on the road surface;

[0104] S111: When it is determined that an abnormal state exists, a real-time warning is issued through a local sound and light alarm device;

[0105] S112: Upload the collected data and logical judgment results to the cloud.

[0106] Preferably, the transport layer module receives data packets from the edge layer module, selects the optimal transmission path according to the destination address of the data packets and the current network environment, and transmits the data packets to the cloud using the project's dedicated network and 4G / 5G dedicated wireless network.

[0107] Preferably, the data layer module includes a natural disaster perception data unit, a video surveillance data unit, a traffic indication status data unit, a traffic safety business supervision data unit and an AI intelligent analysis unit; the data layer module receives data packets from the transport layer module, parses the data packets, and extracts meteorological and natural disaster data; stores the data in a MySQL database, and performs data cleaning and preprocessing. The AI ​​intelligent analysis unit uses a multi-dimensional natural disaster convolutional neural network algorithm to perform deep learning and mining on big data to analyze the trends and patterns of natural disasters.

[0108] Preferably, the data layer module includes the following steps when performing natural disaster data analysis:

[0109] S201: resize the video frame to a uniform size of 96x96 pixels to match the input requirements of the CNN model;

[0110] S202: Convert the image from RGB color space to grayscale;

[0111] S203: Perform normalization processing to scale the pixel value to between 0 and 1;

[0112] S204: labeling images, assigning one or more labels to each image, indicating whether there is a natural disaster and its type;

[0113] S205: aligning the meteorological data, the road geographical environment data and the traffic flow data according to the timestamps;

[0114] S206: Encode the non-numeric data according to the road type and vehicle type, and convert them into a numerical form that can be processed by the model;

[0115] S207: Convert the data into time series and feature vectors;

[0116] S208: Input image data, the input layer receives the pre-processed image frame;

[0117] S209: inputting non-image data, receiving the non-image data through another input layer;

[0118] S210: performing a convolution operation on the input image using multiple convolution kernels to extract features at different levels;

[0119] S211: Each convolution kernel is trained based on natural disaster characteristics;

[0120] S212: Apply ReLU activation function to increase nonlinearity;

[0121] S213: Add a pooling layer after the convolutional layer to reduce the size of the feature map and the amount of computation;

[0122] S214: using a fusion layer to combine the image data and the non-image data by concatenation and weighted summation;

[0123] S215: Flatten the feature maps output by the convolutional layer, pooling layer, and fusion layer, and input them into the fully connected layer;

[0124] S216: The fully connected layer is used to map the extracted features to the target category and output the predicted probability of each category;

[0125] S217: The output layer uses the softmax activation function to output the predicted probability of each natural disaster category;

[0126] As a preferred embodiment, the data layer module also includes the following steps when performing natural disaster data analysis:

[0127] S218: Use the cross entropy loss function to measure the difference between the model prediction and the actual label;

[0128] S219: Select the Adam optimization algorithm to update the model weights and minimize the loss function;

[0129] S220: Divide the labeled data set into a training set, a validation set, and a test set;

[0130] S221: Train the model on the training set and update the weights through back propagation algorithm and gradient descent method;

[0131] S222: Evaluate model performance on the validation set, adjust learning rate, batch size, number of convolution kernels, and optimize the model;

[0132] S223: Use the test set to evaluate the model's accuracy, recall, and F1 score indicators;

[0133] S224: The largest predicted value of each result obtained in S216 is taken as the recognition result, and multiple fully connected layers output the recognition results to form the output result of road natural disaster classification recognition.

[0134] Preferably, the support layer module includes a GIS electronic map, a microservice component, an MQ message queue, a load balancing service unit, a data cache service unit and a security management service unit; the GIS electronic map is used to provide geographic information services for the application layer, and supports map display and path planning; the microservice component is used to split the system into multiple microservices, each of which runs independently; the MQ message queue is used to realize asynchronous communication between modules within the system and improve the concurrent processing capability of the system; the load balancing service unit is used to dynamically allocate requests to different servers according to the system load to achieve load balancing; the data cache service unit is used to cache commonly used data, reduce the access pressure of the database, and improve the response speed of the system; the security management service unit is used to provide security services such as identity authentication, authority management and data encryption.

[0135] Preferably, the application layer modules include a natural disaster real-time monitoring application, a natural disaster early warning alarm application, a natural environment data analysis application and a traffic safety supervision business application; the natural disaster real-time monitoring application is used to display road weather and natural disaster data in real time, and supports map display and data query; the natural disaster early warning alarm application is used to automatically generate early warning alarm information based on the data provided by the data layer and the preset threshold, and publish it through sound and light alarm devices, LED fluorescent display boards, and broadcasting; the natural environment data analysis application is used to conduct in-depth analysis of big data, explore the laws and trends of natural disasters, and provide data support for traffic safety supervision; the traffic safety supervision business application is used to provide decision support for road traffic control or driving induction for the public security traffic police based on the data analysis results.

[0136] Preferably, the security system layer module includes a traffic safety laws, regulations and technical specification system and an information security management and information system level protection system; the traffic safety laws, regulations and technical specification system is used to ensure that the system construction complies with the requirements of national and local laws, regulations and technical specifications; the information security management and information system level protection system ensures the information security and stable operation of the system by formulating information security management systems, strengthening security audits and risk assessments and other measures.

[0137] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge scope of those skilled in the art without departing from the spirit of the present invention.

Claims

1. A road traffic natural disaster early warning and prevention system based on AIOT, characterized by: It includes: The edge layer module is used to collect road meteorological and natural disaster data in real time, perform local preprocessing and logical judgment through edge computing strategies, and conduct preliminary early warning and alarm for abnormal states; The transmission layer module is used to transmit the data collected by the edge layer devices to the cloud server; The data layer module is used to store, clean, analyze and mine the received data, and provide data support for the upper-layer applications; The support layer module is used to provide stable, secure and efficient technical support for the application layer to ensure the efficient operation of the system; The application layer module is used to realize real-time monitoring, early warning, alarm and data analysis of natural disasters based on the data provided by the data layer, and provide decision-making support for traffic safety supervision; The guarantee system layer module is used to adopt the traffic safety laws, regulations and technical specification system and the information security management and information system grading protection system to ensure that the system construction complies with national and local laws and regulations and meets the national information security requirements.

2. According to claim 1, a road traffic natural disaster early warning and prevention system based on AIOT is characterized by: The edge layer module includes a temperature and humidity sensor, a wind direction and speed sensor, a rain sensor, a PM2.5-10 sensor, a video camera and a natural disaster intelligent monitoring terminal; each sensor and the video camera collect road meteorological and natural disaster data in real time, and the natural disaster intelligent monitoring terminal performs preliminary processing on the collected data through edge computing strategies. According to the preset thresholds and logical judgment rules, it judges whether there is an abnormal state of natural disasters, and realizes the early warning and alarm of the abnormal state of the local natural environment through edge computing control strategies.

3. The AIOT-based road traffic natural disaster early warning and prevention system according to claim 2 is characterized by: When the edge layer module is working, it includes the following steps: S101: Each sensor and the video camera collect road meteorological and natural disaster data in real time; S102: The natural disaster intelligent monitoring terminal performs preliminary data cleaning to remove noise data, outliers or duplicate data; S103: Using edge computing strategies, the saturation vapor pressure is calculated by using the formula recommended by Emanuel and the Tetens formula. The formula recommended by Emanuel provided by the World Meteorological Organization is as follows: Among them, E1 is the saturated water vapor pressure, the unit is hPa (hectopascal); T is the absolute temperature scale, and the relationship with Celsius temperature t (℃) is: T = t + 273.15, that is, the relationship between saturated water vapor pressure and Celsius temperature t is as follows: Among them, the Tetens saturation vapor pressure calculation formula has two formulas for water surface and ice surface, which are respectively; Water surface * =6.11×10 7.5t / 237.3+t 、Ice surface E * =6.11×10 9.5t / 265.5+t ; S104: After correcting the Tetens saturation vapor pressure calculation formula, the formula is as follows: Where E2 is the saturated water vapor pressure; t is degrees Celsius; S105: The calculation formulas in steps S104 and S103 are arithmetically averaged to obtain the arithmetic mean value of the saturation vapor pressure. The calculation formula is as follows: S106: According to the definition that relative humidity is the ratio of the actual water vapor pressure in the air to the saturation vapor pressure at the same temperature, the relative humidity is calculated. The calculation formula is as follows: Where E is the actual water vapor pressure in the air at the same temperature, and e is the saturated water vapor pressure at the same temperature; S107: Calculate the actual water vapor pressure according to the relative humidity calculation formula. The calculation formula is as follows: e = f×E; where E is the saturation vapor pressure at the current humidity; S108: Calculate the dew point temperature. The calculation formula is as follows: In the same meteorological environment, when E = e, T1 = t; where T1 is the dew point temperature and t is the monitoring temperature of the temperature sensor; S109: According to the above algorithm derivation, when e = E and t < T1, fog banks form on the road; if the MP2.5-10 value is greater than the threshold, haze forms on the road; if the air temperature continues to drop below 0°C, frost and ice form on the road. S110: By comparing the saturated water vapor pressure with the humidity, it is determined whether the road humidity has reached saturation. If it has reached saturation, the temperature at this time is the dew temperature. If the temperature continues to drop, the water vapor in the air will condense to form "fog". If the PM value is greater than the threshold, "haze" will be formed. At this time, if the temperature continues to drop to 0°C, frost and ice will form on the road surface. S111: When it is determined that an abnormal state exists, a real-time warning is issued through a local sound and light alarm device; S112: Upload the collected data and logical judgment results to the cloud.

4. The AIOT-based road traffic natural disaster early warning and prevention system according to claim 3 is characterized by: The transport layer module receives data packets from the edge layer module, selects the optimal transmission path based on the destination address of the data packet and the current network environment, and transmits the data packet to the cloud using the project's dedicated network and 4G / 5G dedicated wireless network.

5. The AIOT-based road traffic natural disaster early warning and prevention system according to claim 4 is characterized by: The data layer module includes a natural disaster perception data unit, a video surveillance data unit, a traffic indication status data unit, a traffic safety business supervision data unit, and an AI intelligent analysis unit; the data layer module receives data packets from the transport layer module, parses the data packets, and extracts meteorological and natural disaster data; stores the data in the MySQL database, and performs data cleaning and preprocessing. The AI ​​intelligent analysis unit uses a multi-dimensional natural disaster convolutional neural network algorithm to conduct deep learning and mining of big data to analyze the trends and patterns of natural disasters.

6. The AIOT-based road traffic natural disaster early warning and prevention system according to claim 5 is characterized by: The data layer module includes the following steps when analyzing natural disaster data: S201: resize the video frame to a uniform size of 96x96 pixels to match the input requirements of the CNN model; S202: Convert the image from RGB color space to grayscale; S203: Perform normalization processing to scale the pixel value to between 0 and 1; S204: labeling images, assigning one or more labels to each image, indicating whether there is a natural disaster and its type; S205: aligning the meteorological data, the road geographical environment data and the traffic flow data according to the timestamps; S206: Encode the non-numeric data according to the road type and vehicle type, and convert them into a numerical form that can be processed by the model; S207: Convert the data into time series and feature vectors; S208: Input image data, the input layer receives the pre-processed image frame; S209: inputting non-image data, receiving the non-image data through another input layer; S210: performing a convolution operation on the input image using multiple convolution kernels to extract features at different levels; S211: Each convolution kernel is trained based on natural disaster characteristics; S212: Apply ReLU activation function to increase nonlinearity; S213: Add a pooling layer after the convolutional layer to reduce the size of the feature map and the amount of computation; S214: using a fusion layer to combine the image data and the non-image data by concatenation and weighted summation; S215: Flatten the feature maps output by the convolutional layer, pooling layer, and fusion layer, and input them into the fully connected layer; S216: The fully connected layer is used to map the extracted features to the target category and output the predicted probability of each category; S217: The output layer uses the softmax activation function to output the predicted probability of each natural disaster category.

7. The AIOT-based road traffic natural disaster early warning and prevention system according to claim 6 is characterized by: The data layer module also includes the following steps when analyzing natural disaster data: S218: Use the cross entropy loss function to measure the difference between the model prediction and the actual label; S219: Select the Adam optimization algorithm to update the model weights and minimize the loss function; S220: Divide the labeled data set into a training set, a validation set, and a test set; S221: Train the model on the training set and update the weights through back propagation algorithm and gradient descent method; S222: Evaluate model performance on the validation set, adjust learning rate, batch size, number of convolution kernels, and optimize the model; S223: Use the test set to evaluate the model's accuracy, recall, and F1 score indicators; S224: The maximum predicted value of each result obtained in S216 is taken as the recognition result, and multiple fully connected layers output the recognition results to form the output result of road natural disaster classification recognition.

8. The AIOT-based road traffic natural disaster early warning and prevention system according to claim 7 is characterized by: The supporting layer modules include GIS electronic maps, microservice components, MQ message queues, load balancing service units, data cache service units, and security management service units; GIS electronic maps are used to provide geographic information services for the application layer and support map display and path planning; microservice components are used to split the system into multiple microservices, each of which runs independently; MQ message queues are used to implement asynchronous communication between modules within the system and improve the system's concurrent processing capabilities; The load balancing service unit is used to dynamically allocate requests to different servers according to the system load to achieve load balancing; the data cache service unit is used to cache commonly used data, reduce database access pressure, and improve system response speed; the security management service unit is used to provide security services such as identity authentication, permission management, and data encryption.

9. The AIOT-based road traffic natural disaster early warning and prevention system according to claim 8 is characterized by: The application layer modules include real-time monitoring application of natural disasters, early warning and alarm application of natural disasters, natural environment data analysis application and traffic safety supervision business application; the real-time monitoring application of natural disasters is used to display road weather and natural disaster data in real time, and supports map display and data query; the early warning and alarm application of natural disasters is used to automatically generate early warning and alarm information according to the data provided by the data layer and the preset threshold, and publish it through sound and light alarm devices, LED fluorescent display boards, and broadcasting; the natural environment data analysis application is used to conduct in-depth analysis of big data, explore the laws and trends of natural disasters, and provide data support for traffic safety supervision; the traffic safety supervision business application is used to provide decision support for road traffic control or driving induction for the public security traffic police based on the data analysis results.

10. The AIOT-based road traffic natural disaster early warning and prevention system according to claim 9 is characterized by: The security system layer modules include traffic safety laws, regulations and technical specifications system and information security management and information system level protection system; traffic safety laws, regulations and technical specifications system is used to ensure that system construction complies with national and local laws, regulations and technical specifications; information security management and information system level protection system ensures the information security and stable operation of the system by formulating information security management system, strengthening security audit and risk assessment and other measures.

Images